System and Method for Continuous Urban Tree Health Assessment and Predictive Mortality Mapping Using Opportunistic Multi-Pass Spectral Imaging from Municipal Fleet Vehicles and Spatiotemporal Graph Neural Networks
A system that continuously monitors the health of individual urban trees by mounting multispectral cameras (RGB + NIR) on municipal fleet vehicles such as refuse trucks and street sweepers, performing on-device canopy segmentation and spectral index computation, resolving tree identities across multiple vehicle passes, and predicting 6-month mortality risk via a spatiotemporal graph neural network that models spatial relationships between neighboring trees to detect correlated decline from pests, disease, or shared environmental stress.
System and Method for Non-Invasive Sewer Lateral Condition Assessment and Predictive Blockage Detection Using Residential Fixture Drain Acoustic Emission Analysis and Edge-Deployed Spectral Neural Networks
A system that detects developing sewer line blockages and pipe deterioration by capturing acoustic emissions from ordinary residential drain events (toilet flushes, sink usage, shower flow) using a waterproof MEMS microphone at the building cleanout, extracting spectral features including pipe resonance splitting caused by partial obstructions, classifying seven condition states via an edge-deployed CNN, and tracking longitudinal spectral drift to predict time-to-blockage weeks before catastrophic failure.
System and Method for Continuous Indoor Radon Concentration Estimation Without Dedicated Radon Sensors Using Smart Home Barometric Pressure Differential Time-Series, Soil Permeability Transport Models, and Edge-Deployed Physics-Informed Neural Networks
A system that estimates indoor radon concentration without dedicated radon detectors by analyzing barometric pressure differential time-series from existing smart home sensors (thermostats, weather stations, smartphones), computing foundation-level indoor-outdoor pressure differentials accounting for stack effect, wind depressurization, and HVAC mechanical depressurization, and processing these through a physics-informed neural network embedding the Nazaroff advective soil-gas transport equation conditioned on USGS soil radon potential maps, generating continuous radon risk alerts and preemptive ventilation recommendations.
System and Method for Early Wildfire Detection and Ember Transport Corridor Prediction Using Distributed Consumer Weather Station Microclimate Anomaly Fusion and Physics-Informed Neural Networks
A system that fuses barometric pressure microvariations, wind vectors, temperature anomalies, humidity drops, and particulate matter data from the existing network of consumer weather stations to detect wildfire ignition events via spatiotemporal graph neural networks and predict ember transport corridors using physics-informed neural networks embedding Rothermel fire spread and Albini ember lofting models, issuing geofenced alerts to residents in predicted ember landing zones 15–45 minutes before ember arrival.
System and Method for Non-Invasive Continuous Building Envelope Thermal Performance Assessment Using Smart Thermostat Heating and Cooling Response Curve Analysis with Edge-Deployed Machine Learning
A system that continuously infers building envelope thermal resistance and infiltration parameters from heating and cooling response curves already recorded by commodity smart thermostats, cross-referenced with public weather API data, using a Bayesian state-space model to track degradation over months and years and generate actionable alerts for insulation failures, air seal defects, and moisture intrusion events without requiring any additional hardware, sensors, or site visits.
System and Method for Non-Intrusive Residential Water End-Use Disaggregation and Anomalous Flow Detection Using Smart Meter Hydraulic Transient Fingerprinting, Blind Source Separation, and Self-Supervised Contrastive Learning
A system that disaggregates whole-house water consumption into per-fixture end-use categories using hydraulic pressure transient signatures captured by existing smart water meters, applies non-negative tensor factorization for blind source separation of concurrent fixture events, and trains a temporal convolutional network via self-supervised contrastive learning across a utility service territory without labeled data, enabling per-fixture water accounting and real-time leak detection at less than $15 incremental hardware cost per meter.
System and Method for Continuous Road Surface Friction Coefficient Estimation Using Crowdsourced Vehicle Tire-Road Acoustic Emission, Wheel Speed Sensor Micro-Slip Analysis, and Self-Supervised Contrastive Learning for Real-Time Friction Mapping
A system that fuses tire-road acoustic emission captured by existing vehicle cabin microphones with wheel speed sensor micro-slip ratios during normal driving, trains a friction estimation model via self-supervised contrastive learning using cross-vehicle consistency on shared road segments, and aggregates per-vehicle estimates into real-time spatiotemporal friction maps served to navigation and autonomous driving systems, requiring no new hardware on vehicles manufactured after 2020.
System and Method for Real-Time Distribution Transformer Incipient Fault Detection and Remaining Useful Life Estimation Using Smart Meter Voltage Harmonic Spectrum Analysis and Federated Graph Neural Networks
A system that detects incipient distribution transformer faults by analyzing voltage harmonic spectra from existing smart meters using a federated graph neural network that models the radial distribution network topology, distinguishing transformer-originated harmonics from load and feeder sources, enabling per-transformer remaining useful life estimation at zero marginal hardware cost via firmware updates to deployed AMI meters.
System and Method for Real-Time Power Transmission Line Conductor Sag Estimation and Dynamic Line Rating Using Ground-Level Triaxial Magnetometer Arrays and Physics-Informed Recurrent Neural Networks
A system that deploys low-cost triaxial magnetometer arrays at ground level within transmission line rights-of-way, measures the 60 Hz magnetic field gradient produced by overhead conductors, and solves the inverse problem of conductor height estimation using a physics-informed recurrent neural network encoding the Biot-Savart law and IEEE 738 thermal balance equations, enabling real-time dynamic line rating that safely increases transfer capacity by 15–40% at $85 per sensor node with no conductor contact required.
System and Method for Real-Time Urban Flash Flood Depth Prediction and Dynamic Pedestrian-Vehicle Routing Using Distributed Stormwater Level Sensor Networks, LiDAR-Derived Digital Elevation Models, and Physics-Constrained Graph Neural Networks
A system that deploys low-cost ultrasonic water level sensors inside storm drain manholes and catch basins, fuses real-time hydraulic observations with NEXRAD radar rainfall data and municipal pipe network GIS, and uses a physics-constrained graph neural network to predict street-level flood depths 15–60 minutes before overflow occurs, feeding a routing API that dynamically reroutes pedestrians and vehicles around predicted inundation zones.
System and Method for Continuous Crop Water Stress Detection and Precision Irrigation Scheduling via Contactless Monitoring of Plant Xylem Cavitation Acoustic Emissions Using Ultrasonic MEMS Microphone Arrays and Edge-Deployed Neural Networks
A system that detects crop water stress by listening to ultrasonic clicks produced by xylem cavitation events using contactless MEMS microphone arrays positioned near plant stems, classifying stress levels with an on-device temporal convolutional network, and triggering precision irrigation only when and where plants signal physiological need, targeting 20–40% water savings over timer-based or soil-moisture-only systems at $13 per monitored plant.
System and Method for Non-Invasive Refrigerant Charge Estimation in Vapor-Compression HVAC Systems via Compressor Motor Current Signature Analysis and Edge-Deployed Time-Frequency Neural Networks
A system that estimates refrigerant charge level by clamping a $3 split-core current transformer around the compressor power cable and analyzing the motor current waveform's time-frequency signature with an on-device CNN, detecting ±5% charge deviations and fault conditions (liquid slugging, active leaks, valve wear) without any refrigerant-side instrumentation, EPA certification, or system downtime.
System and Method for Continuous Subsurface Void Detection Beneath Roadway Surfaces Using Fleet-Vehicle-Mounted Acoustic Resonance Sensing and Edge-Deployed Spectral Convolutional Neural Networks
A system that repurposes ordinary municipal fleet vehicles (transit buses, refuse trucks, street sweepers) as mobile subsurface void detectors by mounting low-cost MEMS accelerometer arrays and contact microphones to vehicle undercarriages, analyzing the tire-road vibroacoustic coupling signature with an on-device spectral CNN, and aggregating multi-pass observations via Bayesian spatial fusion to build probabilistic city-scale void hazard maps without dedicated survey equipment.
System and Method for Continuous Microplastic Detection and Polymer-Type Classification in Municipal Water Distribution Systems Using Inline Multi-Angle Light Scattering and On-Device Neural Network Particle Identification
An inline flow-through optical cell using a 635 nm laser and eight-angle photodetector array that produces Mie scattering profiles for each particle, classified by an on-device 1D-CNN into polymer type (PE, PP, PS, PET, nylon) and size bin, enabling continuous real-time microplastic monitoring across water distribution networks at under $120 per sensor node.
System and Method for Non-Invasive Buried Pipe Condition Assessment Using Surface-Mounted Vibro-Acoustic Sensor Arrays and Physics-Informed Graph Neural Networks
A non-invasive system that deploys piezoelectric accelerometers at fire hydrants and valve boxes, propagates guided elastic waves through buried pipe walls, and uses a physics-informed graph neural network encoding network topology to infer pipe material, wall thickness, corrosion grade, and remaining useful life without excavation.
System and Method for Continuous Neurodegeneration Screening via Consumer Wearable Gait Microanalysis
An on-device inference pipeline that detects sub-clinical gait asymmetries from consumer IMU data, generating a longitudinal neurodegeneration risk score without cloud connectivity or clinical hardware.
System and Method for Automated Detection of Contradictions in Municipal Code Using Large Language Models
An LLM-based pipeline that ingests complete municipal code corpora, identifies logical contradictions between ordinances, and generates structured conflict reports with citation chains for legislative review.
System and Method for Passive Structural Health Monitoring of Bridges Using Consumer Smartphone Accelerometers and Computer Vision
A crowdsourced system combining smartphone accelerometer data from vehicles crossing bridges with periodic computer-vision crack detection from dashcam footage, generating structural health scores without dedicated sensor installations.
System and Method for Real-Time Pollinator Census in Agricultural Fields Using Edge-Deployed Bioacoustic Classification
A distributed network of low-cost MEMS microphone nodes running on-device audio classification models to identify, count, and map pollinator species activity across agricultural parcels in real time, enabling precision pollination management.
System and Method for Synthetic Grid Inertia Generation Using Coordinated Electric Vehicle Battery Discharge with Predictive Frequency Deviation Modeling
A hierarchical control system that monitors grid frequency in real time and coordinates millisecond-scale power injection from distributed EV batteries to provide synthetic rotational inertia, replacing retiring synchronous generators without dedicated grid-scale storage.
System and Method for Real-Time Multilateral Settlement Netting Across Heterogeneous Payment Rails Using Graph-Based Obligation Compression
A settlement engine that ingests payment obligations from multiple rails (ACH, Fedwire, SWIFT, RTP, stablecoin ledgers), constructs a real-time obligation graph, and computes optimal multilateral netting sets that reduce gross settlement volume by 60-80%, lowering systemic liquidity requirements without requiring participants to share a single ledger.